Meta released its latest AI model, the Muse Spark 1.3, which reportedly stands neck and neck with Anthropic's model, Claude Fable 5.1, and outpaces OpenAI's GPT-5.6 Sol at select metrics such as coding, according to the firm's Chief AI Officer Alexandr Wang, according to a Bloomberg report on Thursday.
The model was made available to developers on Wednesday to access and pay for, with it being called the company's most powerful AI model yet. The update carrying this model will soon be rolled out across the firm's platforms, such as Meta, Facebook and OpenAI.
“This is our biggest jump so far on model performance,” Meta's Chief AI Officer Alexandr Wang stated in an interview on Wednesday, referring to notable progress in the LLM's (Large Language Model) coding and agentic functions as well as its capability to execute tasks in lieu of human users.
Outside of claiming that the Muse Spark 1.3 model stood shoulder-to-shoulder with its peers from OpenAI and Anthropic, Wang also said the model outdoes "any of the current Chinese models out there."
ALSO READ: New AI Models, Critical Threshold, And A Design Debate: A Busy Day For The AI Industry
The article noted that all AI models do certain things better than others and vice-versa. Bloomberg further noted that benchmarking metrics can be gamed and is not necessarily representative of the real world.
Wang stated that the tech giant has not yet decided on releasing the AI model's weights, or the LLM's internal blueprints created during the training cycle, which decides its feedback mechanism. If the weights were to be released to the public, it would enable developers to download, run or build their own AI models using it. The firm is still committed to releasing the weights of its previous build, Muse Spark 1.2, as per Wang.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.